From a7b287bf7845fc335d53550a228ad4fa9aa659e5 Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Wed, 15 May 2019 13:49:25 +0300 Subject: IB/iser: Refactor iscsi_iser_check_protection function Reduce lines of code by using local variable. Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Sagi Grimberg Reviewed-by: Christoph Hellwig Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/iser/iscsi_iser.c | 9 +++------ 1 file changed, 3 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index 9c185a8dabd3..dbad8275b3bc 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -406,13 +406,10 @@ static u8 iscsi_iser_check_protection(struct iscsi_task *task, sector_t *sector) { struct iscsi_iser_task *iser_task = task->dd_data; + enum iser_data_dir dir = iser_task->dir[ISER_DIR_IN] ? + ISER_DIR_IN : ISER_DIR_OUT; - if (iser_task->dir[ISER_DIR_IN]) - return iser_check_task_pi_status(iser_task, ISER_DIR_IN, - sector); - else - return iser_check_task_pi_status(iser_task, ISER_DIR_OUT, - sector); + return iser_check_task_pi_status(iser_task, dir, sector); } /** -- cgit v1.2.3 From 0cc2896f02536272fc58a7d7d2fb2e84f6e717b4 Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Wed, 15 May 2019 13:49:26 +0300 Subject: IB/iser: Remove unused sig_attrs argument Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/iser/iser_memory.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index 2ba70729d7b0..f431c9b4065c 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -302,8 +302,7 @@ void iser_unreg_mem_fastreg(struct iscsi_iser_task *iser_task, } static void -iser_set_dif_domain(struct scsi_cmnd *sc, struct ib_sig_attrs *sig_attrs, - struct ib_sig_domain *domain) +iser_set_dif_domain(struct scsi_cmnd *sc, struct ib_sig_domain *domain) { domain->sig_type = IB_SIG_TYPE_T10_DIF; domain->sig.dif.pi_interval = scsi_prot_interval(sc); @@ -326,21 +325,21 @@ iser_set_sig_attrs(struct scsi_cmnd *sc, struct ib_sig_attrs *sig_attrs) case SCSI_PROT_WRITE_INSERT: case SCSI_PROT_READ_STRIP: sig_attrs->mem.sig_type = IB_SIG_TYPE_NONE; - iser_set_dif_domain(sc, sig_attrs, &sig_attrs->wire); + iser_set_dif_domain(sc, &sig_attrs->wire); sig_attrs->wire.sig.dif.bg_type = IB_T10DIF_CRC; break; case SCSI_PROT_READ_INSERT: case SCSI_PROT_WRITE_STRIP: sig_attrs->wire.sig_type = IB_SIG_TYPE_NONE; - iser_set_dif_domain(sc, sig_attrs, &sig_attrs->mem); + iser_set_dif_domain(sc, &sig_attrs->mem); sig_attrs->mem.sig.dif.bg_type = sc->prot_flags & SCSI_PROT_IP_CHECKSUM ? IB_T10DIF_CSUM : IB_T10DIF_CRC; break; case SCSI_PROT_READ_PASS: case SCSI_PROT_WRITE_PASS: - iser_set_dif_domain(sc, sig_attrs, &sig_attrs->wire); + iser_set_dif_domain(sc, &sig_attrs->wire); sig_attrs->wire.sig.dif.bg_type = IB_T10DIF_CRC; - iser_set_dif_domain(sc, sig_attrs, &sig_attrs->mem); + iser_set_dif_domain(sc, &sig_attrs->mem); sig_attrs->mem.sig.dif.bg_type = sc->prot_flags & SCSI_PROT_IP_CHECKSUM ? IB_T10DIF_CSUM : IB_T10DIF_CRC; break; -- cgit v1.2.3 From dae6345aa7a992fda77b5d0de286c9b238931fbd Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Wed, 15 May 2019 13:49:27 +0300 Subject: IB/isert: Remove unused sig_attrs argument Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/isert/ib_isert.c | 11 +++++------ 1 file changed, 5 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/isert/ib_isert.c b/drivers/infiniband/ulp/isert/ib_isert.c index 989f1ac4245c..ffef4ac152ca 100644 --- a/drivers/infiniband/ulp/isert/ib_isert.c +++ b/drivers/infiniband/ulp/isert/ib_isert.c @@ -2067,8 +2067,7 @@ isert_put_text_rsp(struct iscsi_cmd *cmd, struct iscsi_conn *conn) } static inline void -isert_set_dif_domain(struct se_cmd *se_cmd, struct ib_sig_attrs *sig_attrs, - struct ib_sig_domain *domain) +isert_set_dif_domain(struct se_cmd *se_cmd, struct ib_sig_domain *domain) { domain->sig_type = IB_SIG_TYPE_T10_DIF; domain->sig.dif.bg_type = IB_T10DIF_CRC; @@ -2096,17 +2095,17 @@ isert_set_sig_attrs(struct se_cmd *se_cmd, struct ib_sig_attrs *sig_attrs) case TARGET_PROT_DIN_INSERT: case TARGET_PROT_DOUT_STRIP: sig_attrs->mem.sig_type = IB_SIG_TYPE_NONE; - isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->wire); + isert_set_dif_domain(se_cmd, &sig_attrs->wire); break; case TARGET_PROT_DOUT_INSERT: case TARGET_PROT_DIN_STRIP: sig_attrs->wire.sig_type = IB_SIG_TYPE_NONE; - isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->mem); + isert_set_dif_domain(se_cmd, &sig_attrs->mem); break; case TARGET_PROT_DIN_PASS: case TARGET_PROT_DOUT_PASS: - isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->wire); - isert_set_dif_domain(se_cmd, sig_attrs, &sig_attrs->mem); + isert_set_dif_domain(se_cmd, &sig_attrs->wire); + isert_set_dif_domain(se_cmd, &sig_attrs->mem); break; default: isert_err("Unsupported PI operation %d\n", se_cmd->prot_op); -- cgit v1.2.3 From dfdb08990432b19579703ed5d1efb216b9d2142e Mon Sep 17 00:00:00 2001 From: Leon Romanovsky Date: Mon, 20 May 2019 09:54:20 +0300 Subject: RDMA/ipoib: Remove check of destroy CQ There are nothing to do from user side with knowledge that destroy CQ fails. Signed-off-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/ipoib/ipoib_verbs.c | 7 ++----- 1 file changed, 2 insertions(+), 5 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c index ba09068f6200..b69304d28f06 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_verbs.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_verbs.c @@ -260,11 +260,8 @@ void ipoib_transport_dev_cleanup(struct net_device *dev) priv->qp = NULL; } - if (ib_destroy_cq(priv->send_cq)) - ipoib_warn(priv, "ib_cq_destroy (send) failed\n"); - - if (ib_destroy_cq(priv->recv_cq)) - ipoib_warn(priv, "ib_cq_destroy (recv) failed\n"); + ib_destroy_cq(priv->send_cq); + ib_destroy_cq(priv->recv_cq); } void ipoib_event(struct ib_event_handler *handler, -- cgit v1.2.3 From 633fbb06cbb34cb37117a298c4316ad9d54a7a63 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Wed, 29 May 2019 16:55:45 +0300 Subject: RDMA/ipoib: implement ethtool .get_link() callback Add support for reporting link state for ipoib net devices. $ ip l set dev mlx4_ib0 up $ ethtool mlx4_ib0 | grep Link Link detected: yes $ ip l set dev mlx4_ib0 down $ ethtool mlx4_ib0 | grep Link Link detected: no Signed-off-by: Kamal Heib Reviewed-by: Leon Romanovsky Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/ipoib/ipoib_ethtool.c | 1 + 1 file changed, 1 insertion(+) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c index 83429925dfc6..58016532bf86 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c @@ -222,6 +222,7 @@ static const struct ethtool_ops ipoib_ethtool_ops = { .get_strings = ipoib_get_strings, .get_ethtool_stats = ipoib_get_ethtool_stats, .get_sset_count = ipoib_get_sset_count, + .get_link = ethtool_op_get_link, }; void ipoib_set_ethtool_ops(struct net_device *dev) -- cgit v1.2.3 From bcef5b7215681250c4bf8961dfe15e9e4fef97d0 Mon Sep 17 00:00:00 2001 From: Bart Van Assche Date: Wed, 29 May 2019 09:38:31 -0700 Subject: RDMA/srp: Accept again source addresses that do not have a port number The function srp_parse_in() is used both for parsing source address specifications and for target address specifications. Target addresses must have a port number. Having to specify a port number for source addresses is inconvenient. Make sure that srp_parse_in() supports again parsing addresses with no port number. Cc: Fixes: c62adb7def71 ("IB/srp: Fix IPv6 address parsing") Signed-off-by: Bart Van Assche Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/srp/ib_srp.c | 21 +++++++++++++++------ 1 file changed, 15 insertions(+), 6 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/srp/ib_srp.c b/drivers/infiniband/ulp/srp/ib_srp.c index be9ddcad8f28..87848faa7502 100644 --- a/drivers/infiniband/ulp/srp/ib_srp.c +++ b/drivers/infiniband/ulp/srp/ib_srp.c @@ -3481,13 +3481,14 @@ static const match_table_t srp_opt_tokens = { * @net: [in] Network namespace. * @sa: [out] Address family, IP address and port number. * @addr_port_str: [in] IP address and port number. + * @has_port: [out] Whether or not @addr_port_str includes a port number. * * Parse the following address formats: * - IPv4: :, e.g. 1.2.3.4:5. * - IPv6: \[\]:, e.g. [1::2:3%4]:5. */ static int srp_parse_in(struct net *net, struct sockaddr_storage *sa, - const char *addr_port_str) + const char *addr_port_str, bool *has_port) { char *addr_end, *addr = kstrdup(addr_port_str, GFP_KERNEL); char *port_str; @@ -3496,9 +3497,12 @@ static int srp_parse_in(struct net *net, struct sockaddr_storage *sa, if (!addr) return -ENOMEM; port_str = strrchr(addr, ':'); - if (!port_str) - return -EINVAL; - *port_str++ = '\0'; + if (port_str && strchr(port_str, ']')) + port_str = NULL; + if (port_str) + *port_str++ = '\0'; + if (has_port) + *has_port = port_str != NULL; ret = inet_pton_with_scope(net, AF_INET, addr, port_str, sa); if (ret && addr[0]) { addr_end = addr + strlen(addr) - 1; @@ -3520,6 +3524,7 @@ static int srp_parse_options(struct net *net, const char *buf, char *p; substring_t args[MAX_OPT_ARGS]; unsigned long long ull; + bool has_port; int opt_mask = 0; int token; int ret = -EINVAL; @@ -3618,7 +3623,8 @@ static int srp_parse_options(struct net *net, const char *buf, ret = -ENOMEM; goto out; } - ret = srp_parse_in(net, &target->rdma_cm.src.ss, p); + ret = srp_parse_in(net, &target->rdma_cm.src.ss, p, + NULL); if (ret < 0) { pr_warn("bad source parameter '%s'\n", p); kfree(p); @@ -3634,7 +3640,10 @@ static int srp_parse_options(struct net *net, const char *buf, ret = -ENOMEM; goto out; } - ret = srp_parse_in(net, &target->rdma_cm.dst.ss, p); + ret = srp_parse_in(net, &target->rdma_cm.dst.ss, p, + &has_port); + if (!has_port) + ret = -EINVAL; if (ret < 0) { pr_warn("bad dest parameter '%s'\n", p); kfree(p); -- cgit v1.2.3 From cbdc666f3e842b01d4537933e0a64f1e7cf17017 Mon Sep 17 00:00:00 2001 From: Kamal Heib Date: Thu, 30 May 2019 16:18:17 +0300 Subject: RDMA/ipoib: Remove check for ETH_SS_TEST The default action for unlisted tests is "not-supported", so given that ipoib doesn't support ETH_SS_TEST, there is no need to check for it in the case statements, just let it get caught by the default: case. Fixes: e3614bc9dc44 ("IB/ipoib: Add readout of statistics using ethtool") Signed-off-by: Kamal Heib Signed-off-by: Doug Ledford --- drivers/infiniband/ulp/ipoib/ipoib_ethtool.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c index 58016532bf86..63e4f9d15fd9 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_ethtool.c @@ -138,7 +138,6 @@ static void ipoib_get_strings(struct net_device __always_unused *dev, p += ETH_GSTRING_LEN; } break; - case ETH_SS_TEST: default: break; } @@ -149,7 +148,6 @@ static int ipoib_get_sset_count(struct net_device __always_unused *dev, switch (sset) { case ETH_SS_STATS: return IPOIB_GLOBAL_STATS_LEN; - case ETH_SS_TEST: default: break; } -- cgit v1.2.3 From b76a439982f8483beb2ffcfe93be50026940030a Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Tue, 11 Jun 2019 18:52:47 +0300 Subject: IB/iser: Use IB_WR_REG_MR_INTEGRITY for PI handover Using this new API reduces iSER code complexity. It also reduces the maximum number of work requests per task and the need of dealing with multiple MRs (and their registrations and invalidations) per task. It is done by using a single WR and a special MR type (IB_MR_TYPE_INTEGRITY) for PI operation. The setup of the tested benchmark: - 2 servers with 24 cores (1 initiator and 1 target) - 24 target sessions with 1 LUN each - ramdisk backstore - PI active Performance results running fio (24 jobs, 128 iodepth) using write_generate=0 and read_verify=0 (w/w.o patch): bs IOPS(read) IOPS(write) ---- ---------- ---------- 512 1236.6K/1164.3K 1357.2K/1332.8K 1k 1196.5K/1163.8K 1348.4K/1262.7K 2k 1016.7K/921950 1003.7K/931230 4k 662728/600545 595423/501513 8k 385954/384345 333775/277090 16k 222864/222820 170317/170671 32k 116869/114896 82331/82244 64k 55205/54931 40264/40021 Using write_generate=1 and read_verify=1 (w/w.o patch): bs IOPS(read) IOPS(write) ---- ---------- ---------- 512 1090.1K/1030.9K 1303.9K/1101.4K 1k 1057.7K/904583 1318.4K/988085 2k 965226/638799 1008.6K/692514 4k 555479/410151 542414/414517 8k 298675/224964 264729/237508 16k 133485/122481 164625/138647 32k 74329/67615 80143/78743 64k 35716/35519 39294/37334 We get performance improvement at all block sizes. The most significant improvement is when writing 4k bs (almost 30% more iops). Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/iser/iscsi_iser.h | 38 ++------ drivers/infiniband/ulp/iser/iser_initiator.c | 12 ++- drivers/infiniband/ulp/iser/iser_memory.c | 98 +++++++------------ drivers/infiniband/ulp/iser/iser_verbs.c | 140 +++++++++------------------ 4 files changed, 95 insertions(+), 193 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 36d525110fd2..6bf9eaa8ec96 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -225,13 +225,11 @@ enum iser_desc_type { ISCSI_TX_DATAOUT }; -/* Maximum number of work requests per task: - * Data memory region local invalidate + fast registration - * Protection memory region local invalidate + fast registration - * Signature memory region local invalidate + fast registration - * PDU send +/* + * Maximum number of work requests per task + * (invalidate, registration, send) */ -#define ISER_MAX_WRS 7 +#define ISER_MAX_WRS 3 /** * struct iser_tx_desc - iSER TX descriptor @@ -247,9 +245,6 @@ enum iser_desc_type { * @mapped: Is the task header mapped * @wr_idx: Current WR index * @wrs: Array of WRs per task - * @data_reg: Data buffer registration details - * @prot_reg: Protection buffer registration details - * @sig_attrs: Signature attributes */ struct iser_tx_desc { struct iser_ctrl iser_header; @@ -264,11 +259,7 @@ struct iser_tx_desc { union iser_wr { struct ib_send_wr send; struct ib_reg_wr fast_reg; - struct ib_sig_handover_wr sig; } wrs[ISER_MAX_WRS]; - struct iser_mem_reg data_reg; - struct iser_mem_reg prot_reg; - struct ib_sig_attrs sig_attrs; }; #define ISER_RX_PAD_SIZE (256 - (ISER_RX_PAYLOAD_SIZE + \ @@ -388,6 +379,7 @@ struct iser_device { * * @mr: memory region * @fmr_pool: pool of fmrs + * @sig_mr: signature memory region * @page_vec: fast reg page list used by fmr pool * @mr_valid: is mr valid indicator */ @@ -396,36 +388,22 @@ struct iser_reg_resources { struct ib_mr *mr; struct ib_fmr_pool *fmr_pool; }; + struct ib_mr *sig_mr; struct iser_page_vec *page_vec; u8 mr_valid:1; }; -/** - * struct iser_pi_context - Protection information context - * - * @rsc: protection buffer registration resources - * @sig_mr: signature enable memory region - * @sig_mr_valid: is sig_mr valid indicator - * @sig_protected: is region protected indicator - */ -struct iser_pi_context { - struct iser_reg_resources rsc; - struct ib_mr *sig_mr; - u8 sig_mr_valid:1; - u8 sig_protected:1; -}; - /** * struct iser_fr_desc - Fast registration descriptor * * @list: entry in connection fastreg pool * @rsc: data buffer registration resources - * @pi_ctx: protection information context + * @sig_protected: is region protected indicator */ struct iser_fr_desc { struct list_head list; struct iser_reg_resources rsc; - struct iser_pi_context *pi_ctx; + bool sig_protected; struct list_head all_list; }; diff --git a/drivers/infiniband/ulp/iser/iser_initiator.c b/drivers/infiniband/ulp/iser/iser_initiator.c index 96af06cfe0af..5cbb4b3a0566 100644 --- a/drivers/infiniband/ulp/iser/iser_initiator.c +++ b/drivers/infiniband/ulp/iser/iser_initiator.c @@ -592,15 +592,14 @@ void iser_login_rsp(struct ib_cq *cq, struct ib_wc *wc) static inline int iser_inv_desc(struct iser_fr_desc *desc, u32 rkey) { - if (likely(rkey == desc->rsc.mr->rkey)) { - desc->rsc.mr_valid = 0; - } else if (likely(desc->pi_ctx && rkey == desc->pi_ctx->sig_mr->rkey)) { - desc->pi_ctx->sig_mr_valid = 0; - } else { + if (unlikely((!desc->sig_protected && rkey != desc->rsc.mr->rkey) || + (desc->sig_protected && rkey != desc->rsc.sig_mr->rkey))) { iser_err("Bogus remote invalidation for rkey %#x\n", rkey); return -EINVAL; } + desc->rsc.mr_valid = 0; + return 0; } @@ -750,6 +749,9 @@ void iser_task_rdma_init(struct iscsi_iser_task *iser_task) iser_task->prot[ISER_DIR_IN].data_len = 0; iser_task->prot[ISER_DIR_OUT].data_len = 0; + iser_task->prot[ISER_DIR_IN].dma_nents = 0; + iser_task->prot[ISER_DIR_OUT].dma_nents = 0; + memset(&iser_task->rdma_reg[ISER_DIR_IN], 0, sizeof(struct iser_mem_reg)); memset(&iser_task->rdma_reg[ISER_DIR_OUT], 0, diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index f431c9b4065c..d66e17c2a085 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -376,16 +376,16 @@ iser_inv_rkey(struct ib_send_wr *inv_wr, static int iser_reg_sig_mr(struct iscsi_iser_task *iser_task, - struct iser_pi_context *pi_ctx, - struct iser_mem_reg *data_reg, - struct iser_mem_reg *prot_reg, + struct iser_data_buf *mem, + struct iser_data_buf *sig_mem, + struct iser_reg_resources *rsc, struct iser_mem_reg *sig_reg) { struct iser_tx_desc *tx_desc = &iser_task->desc; - struct ib_sig_attrs *sig_attrs = &tx_desc->sig_attrs; struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe; - struct ib_sig_handover_wr *wr; - struct ib_mr *mr = pi_ctx->sig_mr; + struct ib_mr *mr = rsc->sig_mr; + struct ib_sig_attrs *sig_attrs = mr->sig_attrs; + struct ib_reg_wr *wr; int ret; memset(sig_attrs, 0, sizeof(*sig_attrs)); @@ -395,33 +395,36 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, iser_set_prot_checks(iser_task->sc, &sig_attrs->check_mask); - if (pi_ctx->sig_mr_valid) + if (rsc->mr_valid) iser_inv_rkey(iser_tx_next_wr(tx_desc), mr, cqe); ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey)); - wr = container_of(iser_tx_next_wr(tx_desc), struct ib_sig_handover_wr, - wr); - wr->wr.opcode = IB_WR_REG_SIG_MR; + ret = ib_map_mr_sg_pi(mr, mem->sg, mem->dma_nents, NULL, + sig_mem->sg, sig_mem->dma_nents, NULL, SZ_4K); + if (unlikely(ret)) { + iser_err("failed to map PI sg (%d)\n", + mem->dma_nents + sig_mem->dma_nents); + goto err; + } + + wr = container_of(iser_tx_next_wr(tx_desc), struct ib_reg_wr, wr); + memset(wr, 0, sizeof(*wr)); + wr->wr.opcode = IB_WR_REG_MR_INTEGRITY; wr->wr.wr_cqe = cqe; - wr->wr.sg_list = &data_reg->sge; - wr->wr.num_sge = 1; + wr->wr.num_sge = 0; wr->wr.send_flags = 0; - wr->sig_attrs = sig_attrs; - wr->sig_mr = mr; - if (scsi_prot_sg_count(iser_task->sc)) - wr->prot = &prot_reg->sge; - else - wr->prot = NULL; - wr->access_flags = IB_ACCESS_LOCAL_WRITE | - IB_ACCESS_REMOTE_READ | - IB_ACCESS_REMOTE_WRITE; - pi_ctx->sig_mr_valid = 1; + wr->mr = mr; + wr->key = mr->rkey; + wr->access = IB_ACCESS_LOCAL_WRITE | + IB_ACCESS_REMOTE_READ | + IB_ACCESS_REMOTE_WRITE; + rsc->mr_valid = 1; sig_reg->sge.lkey = mr->lkey; sig_reg->rkey = mr->rkey; - sig_reg->sge.addr = 0; - sig_reg->sge.length = scsi_transfer_length(iser_task->sc); + sig_reg->sge.addr = mr->iova; + sig_reg->sge.length = mr->length; iser_dbg("lkey=0x%x rkey=0x%x addr=0x%llx length=%u\n", sig_reg->sge.lkey, sig_reg->rkey, sig_reg->sge.addr, @@ -477,21 +480,6 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, return 0; } -static int -iser_reg_prot_sg(struct iscsi_iser_task *task, - struct iser_data_buf *mem, - struct iser_fr_desc *desc, - bool use_dma_key, - struct iser_mem_reg *reg) -{ - struct iser_device *device = task->iser_conn->ib_conn.device; - - if (use_dma_key) - return iser_reg_dma(device, mem, reg); - - return device->reg_ops->reg_mem(task, mem, &desc->pi_ctx->rsc, reg); -} - static int iser_reg_data_sg(struct iscsi_iser_task *task, struct iser_data_buf *mem, @@ -515,7 +503,6 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task, struct iser_device *device = ib_conn->device; struct iser_data_buf *mem = &task->data[dir]; struct iser_mem_reg *reg = &task->rdma_reg[dir]; - struct iser_mem_reg *data_reg; struct iser_fr_desc *desc = NULL; bool use_dma_key; int err; @@ -528,32 +515,17 @@ int iser_reg_rdma_mem(struct iscsi_iser_task *task, reg->mem_h = desc; } - if (scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL) - data_reg = reg; - else - data_reg = &task->desc.data_reg; - - err = iser_reg_data_sg(task, mem, desc, use_dma_key, data_reg); - if (unlikely(err)) - goto err_reg; - - if (scsi_get_prot_op(task->sc) != SCSI_PROT_NORMAL) { - struct iser_mem_reg *prot_reg = &task->desc.prot_reg; - - if (scsi_prot_sg_count(task->sc)) { - mem = &task->prot[dir]; - err = iser_reg_prot_sg(task, mem, desc, - use_dma_key, prot_reg); - if (unlikely(err)) - goto err_reg; - } - - err = iser_reg_sig_mr(task, desc->pi_ctx, data_reg, - prot_reg, reg); + if (scsi_get_prot_op(task->sc) == SCSI_PROT_NORMAL) { + err = iser_reg_data_sg(task, mem, desc, use_dma_key, reg); + if (unlikely(err)) + goto err_reg; + } else { + err = iser_reg_sig_mr(task, mem, &task->prot[dir], + &desc->rsc, reg); if (unlikely(err)) goto err_reg; - desc->pi_ctx->sig_protected = 1; + desc->sig_protected = 1; } return 0; diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index 4ff3d98fa6a4..ffd6bbc819f7 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -233,116 +233,63 @@ void iser_free_fmr_pool(struct ib_conn *ib_conn) kfree(desc); } -static int -iser_alloc_reg_res(struct iser_device *device, - struct ib_pd *pd, - struct iser_reg_resources *res, - unsigned int size) +static struct iser_fr_desc * +iser_create_fastreg_desc(struct iser_device *device, + struct ib_pd *pd, + bool pi_enable, + unsigned int size) { + struct iser_fr_desc *desc; struct ib_device *ib_dev = device->ib_device; enum ib_mr_type mr_type; int ret; + desc = kzalloc(sizeof(*desc), GFP_KERNEL); + if (!desc) + return ERR_PTR(-ENOMEM); + if (ib_dev->attrs.device_cap_flags & IB_DEVICE_SG_GAPS_REG) mr_type = IB_MR_TYPE_SG_GAPS; else mr_type = IB_MR_TYPE_MEM_REG; - res->mr = ib_alloc_mr(pd, mr_type, size); - if (IS_ERR(res->mr)) { - ret = PTR_ERR(res->mr); + desc->rsc.mr = ib_alloc_mr(pd, mr_type, size); + if (IS_ERR(desc->rsc.mr)) { + ret = PTR_ERR(desc->rsc.mr); iser_err("Failed to allocate ib_fast_reg_mr err=%d\n", ret); - return ret; - } - res->mr_valid = 0; - - return 0; -} - -static void -iser_free_reg_res(struct iser_reg_resources *rsc) -{ - ib_dereg_mr(rsc->mr); -} - -static int -iser_alloc_pi_ctx(struct iser_device *device, - struct ib_pd *pd, - struct iser_fr_desc *desc, - unsigned int size) -{ - struct iser_pi_context *pi_ctx = NULL; - int ret; - - desc->pi_ctx = kzalloc(sizeof(*desc->pi_ctx), GFP_KERNEL); - if (!desc->pi_ctx) - return -ENOMEM; - - pi_ctx = desc->pi_ctx; - - ret = iser_alloc_reg_res(device, pd, &pi_ctx->rsc, size); - if (ret) { - iser_err("failed to allocate reg_resources\n"); - goto alloc_reg_res_err; + goto err_alloc_mr; } - pi_ctx->sig_mr = ib_alloc_mr(pd, IB_MR_TYPE_SIGNATURE, 2); - if (IS_ERR(pi_ctx->sig_mr)) { - ret = PTR_ERR(pi_ctx->sig_mr); - goto sig_mr_failure; + if (pi_enable) { + desc->rsc.sig_mr = ib_alloc_mr_integrity(pd, size, size); + if (IS_ERR(desc->rsc.sig_mr)) { + ret = PTR_ERR(desc->rsc.sig_mr); + iser_err("Failed to allocate sig_mr err=%d\n", ret); + goto err_alloc_mr_integrity; + } } - pi_ctx->sig_mr_valid = 0; - desc->pi_ctx->sig_protected = 0; + desc->rsc.mr_valid = 0; - return 0; + return desc; -sig_mr_failure: - iser_free_reg_res(&pi_ctx->rsc); -alloc_reg_res_err: - kfree(desc->pi_ctx); +err_alloc_mr_integrity: + ib_dereg_mr(desc->rsc.mr); +err_alloc_mr: + kfree(desc); - return ret; + return ERR_PTR(ret); } -static void -iser_free_pi_ctx(struct iser_pi_context *pi_ctx) +static void iser_destroy_fastreg_desc(struct iser_fr_desc *desc) { - iser_free_reg_res(&pi_ctx->rsc); - ib_dereg_mr(pi_ctx->sig_mr); - kfree(pi_ctx); -} - -static struct iser_fr_desc * -iser_create_fastreg_desc(struct iser_device *device, - struct ib_pd *pd, - bool pi_enable, - unsigned int size) -{ - struct iser_fr_desc *desc; - int ret; + struct iser_reg_resources *res = &desc->rsc; - desc = kzalloc(sizeof(*desc), GFP_KERNEL); - if (!desc) - return ERR_PTR(-ENOMEM); - - ret = iser_alloc_reg_res(device, pd, &desc->rsc, size); - if (ret) - goto reg_res_alloc_failure; - - if (pi_enable) { - ret = iser_alloc_pi_ctx(device, pd, desc, size); - if (ret) - goto pi_ctx_alloc_failure; + ib_dereg_mr(res->mr); + if (res->sig_mr) { + ib_dereg_mr(res->sig_mr); + res->sig_mr = NULL; } - - return desc; - -pi_ctx_alloc_failure: - iser_free_reg_res(&desc->rsc); -reg_res_alloc_failure: kfree(desc); - - return ERR_PTR(ret); } /** @@ -399,10 +346,7 @@ void iser_free_fastreg_pool(struct ib_conn *ib_conn) list_for_each_entry_safe(desc, tmp, &fr_pool->all_list, all_list) { list_del(&desc->all_list); - iser_free_reg_res(&desc->rsc); - if (desc->pi_ctx) - iser_free_pi_ctx(desc->pi_ctx); - kfree(desc); + iser_destroy_fastreg_desc(desc); ++i; } @@ -707,6 +651,7 @@ iser_calc_scsi_params(struct iser_conn *iser_conn, struct ib_device_attr *attr = &device->ib_device->attrs; unsigned short sg_tablesize, sup_sg_tablesize; unsigned short reserved_mr_pages; + u32 max_num_sg; /* * FRs without SG_GAPS or FMRs can only map up to a (device) page per @@ -720,12 +665,17 @@ iser_calc_scsi_params(struct iser_conn *iser_conn, else reserved_mr_pages = 1; + if (iser_conn->ib_conn.pi_support) + max_num_sg = attr->max_pi_fast_reg_page_list_len; + else + max_num_sg = attr->max_fast_reg_page_list_len; + sg_tablesize = DIV_ROUND_UP(max_sectors * 512, SIZE_4K); if (attr->device_cap_flags & IB_DEVICE_MEM_MGT_EXTENSIONS) sup_sg_tablesize = min_t( uint, ISCSI_ISER_MAX_SG_TABLESIZE, - attr->max_fast_reg_page_list_len - reserved_mr_pages); + max_num_sg - reserved_mr_pages); else sup_sg_tablesize = ISCSI_ISER_MAX_SG_TABLESIZE; @@ -1118,9 +1068,9 @@ u8 iser_check_task_pi_status(struct iscsi_iser_task *iser_task, struct ib_mr_status mr_status; int ret; - if (desc && desc->pi_ctx->sig_protected) { - desc->pi_ctx->sig_protected = 0; - ret = ib_check_mr_status(desc->pi_ctx->sig_mr, + if (desc && desc->sig_protected) { + desc->sig_protected = 0; + ret = ib_check_mr_status(desc->rsc.sig_mr, IB_MR_CHECK_SIG_STATUS, &mr_status); if (ret) { pr_err("ib_check_mr_status failed, ret %d\n", ret); -- cgit v1.2.3 From b9294f8b7c4bfdad35f4eb8330974892aedd1aaf Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Tue, 11 Jun 2019 18:52:48 +0300 Subject: IB/iser: Unwind WR union at iser_tx_desc After decreasing WRs array size from 7 to 3 it is more readable to give each WR a descriptive name. Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig Reviewed-by: Sagi Grimberg Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/iser/iscsi_iser.c | 3 ++- drivers/infiniband/ulp/iser/iscsi_iser.h | 34 ++++++------------------------- drivers/infiniband/ulp/iser/iser_memory.c | 16 ++++++++------- drivers/infiniband/ulp/iser/iser_verbs.c | 12 +++++++++-- 4 files changed, 27 insertions(+), 38 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.c b/drivers/infiniband/ulp/iser/iscsi_iser.c index dbad8275b3bc..c7a3d75fb308 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.c +++ b/drivers/infiniband/ulp/iser/iscsi_iser.c @@ -205,7 +205,8 @@ iser_initialize_task_headers(struct iscsi_task *task, goto out; } - tx_desc->wr_idx = 0; + tx_desc->inv_wr.next = NULL; + tx_desc->reg_wr.wr.next = NULL; tx_desc->mapped = true; tx_desc->dma_addr = dma_addr; tx_desc->tx_sg[0].addr = tx_desc->dma_addr; diff --git a/drivers/infiniband/ulp/iser/iscsi_iser.h b/drivers/infiniband/ulp/iser/iscsi_iser.h index 6bf9eaa8ec96..39bf213444cb 100644 --- a/drivers/infiniband/ulp/iser/iscsi_iser.h +++ b/drivers/infiniband/ulp/iser/iscsi_iser.h @@ -225,12 +225,6 @@ enum iser_desc_type { ISCSI_TX_DATAOUT }; -/* - * Maximum number of work requests per task - * (invalidate, registration, send) - */ -#define ISER_MAX_WRS 3 - /** * struct iser_tx_desc - iSER TX descriptor * @@ -243,8 +237,9 @@ enum iser_desc_type { * unsolicited data-out or control * @num_sge: number sges used on this TX task * @mapped: Is the task header mapped - * @wr_idx: Current WR index - * @wrs: Array of WRs per task + * reg_wr: registration WR + * send_wr: send WR + * inv_wr: invalidate WR */ struct iser_tx_desc { struct iser_ctrl iser_header; @@ -255,11 +250,9 @@ struct iser_tx_desc { int num_sge; struct ib_cqe cqe; bool mapped; - u8 wr_idx; - union iser_wr { - struct ib_send_wr send; - struct ib_reg_wr fast_reg; - } wrs[ISER_MAX_WRS]; + struct ib_reg_wr reg_wr; + struct ib_send_wr send_wr; + struct ib_send_wr inv_wr; }; #define ISER_RX_PAD_SIZE (256 - (ISER_RX_PAYLOAD_SIZE + \ @@ -652,21 +645,6 @@ void iser_reg_desc_put_fmr(struct ib_conn *ib_conn, struct iser_fr_desc *desc); -static inline struct ib_send_wr * -iser_tx_next_wr(struct iser_tx_desc *tx_desc) -{ - struct ib_send_wr *cur_wr = &tx_desc->wrs[tx_desc->wr_idx].send; - struct ib_send_wr *last_wr; - - if (tx_desc->wr_idx) { - last_wr = &tx_desc->wrs[tx_desc->wr_idx - 1].send; - last_wr->next = cur_wr; - } - tx_desc->wr_idx++; - - return cur_wr; -} - static inline struct iser_conn * to_iser_conn(struct ib_conn *ib_conn) { diff --git a/drivers/infiniband/ulp/iser/iser_memory.c b/drivers/infiniband/ulp/iser/iser_memory.c index d66e17c2a085..2cc89a9b9e9b 100644 --- a/drivers/infiniband/ulp/iser/iser_memory.c +++ b/drivers/infiniband/ulp/iser/iser_memory.c @@ -365,13 +365,15 @@ iser_set_prot_checks(struct scsi_cmnd *sc, u8 *mask) static inline void iser_inv_rkey(struct ib_send_wr *inv_wr, struct ib_mr *mr, - struct ib_cqe *cqe) + struct ib_cqe *cqe, + struct ib_send_wr *next_wr) { inv_wr->opcode = IB_WR_LOCAL_INV; inv_wr->wr_cqe = cqe; inv_wr->ex.invalidate_rkey = mr->rkey; inv_wr->send_flags = 0; inv_wr->num_sge = 0; + inv_wr->next = next_wr; } static int @@ -385,7 +387,7 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe; struct ib_mr *mr = rsc->sig_mr; struct ib_sig_attrs *sig_attrs = mr->sig_attrs; - struct ib_reg_wr *wr; + struct ib_reg_wr *wr = &tx_desc->reg_wr; int ret; memset(sig_attrs, 0, sizeof(*sig_attrs)); @@ -396,7 +398,7 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, iser_set_prot_checks(iser_task->sc, &sig_attrs->check_mask); if (rsc->mr_valid) - iser_inv_rkey(iser_tx_next_wr(tx_desc), mr, cqe); + iser_inv_rkey(&tx_desc->inv_wr, mr, cqe, &wr->wr); ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey)); @@ -408,8 +410,8 @@ iser_reg_sig_mr(struct iscsi_iser_task *iser_task, goto err; } - wr = container_of(iser_tx_next_wr(tx_desc), struct ib_reg_wr, wr); memset(wr, 0, sizeof(*wr)); + wr->wr.next = &tx_desc->send_wr; wr->wr.opcode = IB_WR_REG_MR_INTEGRITY; wr->wr.wr_cqe = cqe; wr->wr.num_sge = 0; @@ -441,11 +443,11 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, struct iser_tx_desc *tx_desc = &iser_task->desc; struct ib_cqe *cqe = &iser_task->iser_conn->ib_conn.reg_cqe; struct ib_mr *mr = rsc->mr; - struct ib_reg_wr *wr; + struct ib_reg_wr *wr = &tx_desc->reg_wr; int n; if (rsc->mr_valid) - iser_inv_rkey(iser_tx_next_wr(tx_desc), mr, cqe); + iser_inv_rkey(&tx_desc->inv_wr, mr, cqe, &wr->wr); ib_update_fast_reg_key(mr, ib_inc_rkey(mr->rkey)); @@ -456,7 +458,7 @@ static int iser_fast_reg_mr(struct iscsi_iser_task *iser_task, return n < 0 ? n : -EINVAL; } - wr = container_of(iser_tx_next_wr(tx_desc), struct ib_reg_wr, wr); + wr->wr.next = &tx_desc->send_wr; wr->wr.opcode = IB_WR_REG_MR; wr->wr.wr_cqe = cqe; wr->wr.send_flags = 0; diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index ffd6bbc819f7..ea9cf04ad002 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -1037,7 +1037,8 @@ int iser_post_recvm(struct iser_conn *iser_conn, int count) int iser_post_send(struct ib_conn *ib_conn, struct iser_tx_desc *tx_desc, bool signal) { - struct ib_send_wr *wr = iser_tx_next_wr(tx_desc); + struct ib_send_wr *wr = &tx_desc->send_wr; + struct ib_send_wr *first_wr; int ib_ret; ib_dma_sync_single_for_device(ib_conn->device->ib_device, @@ -1051,7 +1052,14 @@ int iser_post_send(struct ib_conn *ib_conn, struct iser_tx_desc *tx_desc, wr->opcode = IB_WR_SEND; wr->send_flags = signal ? IB_SEND_SIGNALED : 0; - ib_ret = ib_post_send(ib_conn->qp, &tx_desc->wrs[0].send, NULL); + if (tx_desc->inv_wr.next) + first_wr = &tx_desc->inv_wr; + else if (tx_desc->reg_wr.wr.next) + first_wr = &tx_desc->reg_wr.wr; + else + first_wr = wr; + + ib_ret = ib_post_send(ib_conn->qp, first_wr, NULL); if (ib_ret) iser_err("ib_post_send failed, ret:%d opcode:%d\n", ib_ret, wr->opcode); -- cgit v1.2.3 From c0a6cbb9cbccffc249743afa16e64f16c46c80b2 Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Tue, 11 Jun 2019 18:52:50 +0300 Subject: RDMA/core: Rename signature qp create flag and signature device capability Rename IB_QP_CREATE_SIGNATURE_EN to IB_QP_CREATE_INTEGRITY_EN and IB_DEVICE_SIGNATURE_HANDOVER to IB_DEVICE_INTEGRITY_HANDOVER. Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/iser/iser_verbs.c | 4 ++-- drivers/infiniband/ulp/isert/ib_isert.c | 4 ++-- 2 files changed, 4 insertions(+), 4 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/iser/iser_verbs.c b/drivers/infiniband/ulp/iser/iser_verbs.c index ea9cf04ad002..a6548de0e218 100644 --- a/drivers/infiniband/ulp/iser/iser_verbs.c +++ b/drivers/infiniband/ulp/iser/iser_verbs.c @@ -399,7 +399,7 @@ static int iser_create_ib_conn_res(struct ib_conn *ib_conn) init_attr.qp_type = IB_QPT_RC; if (ib_conn->pi_support) { init_attr.cap.max_send_wr = ISER_QP_SIG_MAX_REQ_DTOS + 1; - init_attr.create_flags |= IB_QP_CREATE_SIGNATURE_EN; + init_attr.create_flags |= IB_QP_CREATE_INTEGRITY_EN; iser_conn->max_cmds = ISER_GET_MAX_XMIT_CMDS(ISER_QP_SIG_MAX_REQ_DTOS); } else { @@ -712,7 +712,7 @@ static void iser_addr_handler(struct rdma_cm_id *cma_id) /* connection T10-PI support */ if (iser_pi_enable) { if (!(device->ib_device->attrs.device_cap_flags & - IB_DEVICE_SIGNATURE_HANDOVER)) { + IB_DEVICE_INTEGRITY_HANDOVER)) { iser_warn("T10-PI requested but not supported on %s, " "continue without T10-PI\n", dev_name(&ib_conn->device->ib_device->dev)); diff --git a/drivers/infiniband/ulp/isert/ib_isert.c b/drivers/infiniband/ulp/isert/ib_isert.c index ffef4ac152ca..53bc505f5292 100644 --- a/drivers/infiniband/ulp/isert/ib_isert.c +++ b/drivers/infiniband/ulp/isert/ib_isert.c @@ -141,7 +141,7 @@ isert_create_qp(struct isert_conn *isert_conn, attr.sq_sig_type = IB_SIGNAL_REQ_WR; attr.qp_type = IB_QPT_RC; if (device->pi_capable) - attr.create_flags |= IB_QP_CREATE_SIGNATURE_EN; + attr.create_flags |= IB_QP_CREATE_INTEGRITY_EN; ret = rdma_create_qp(cma_id, device->pd, &attr); if (ret) { @@ -317,7 +317,7 @@ isert_create_device_ib_res(struct isert_device *device) /* Check signature cap */ device->pi_capable = ib_dev->attrs.device_cap_flags & - IB_DEVICE_SIGNATURE_HANDOVER ? true : false; + IB_DEVICE_INTEGRITY_HANDOVER ? true : false; return 0; -- cgit v1.2.3 From e9a53e73a2507f3a1680538bd167b2e49533659a Mon Sep 17 00:00:00 2001 From: Israel Rukshin Date: Tue, 11 Jun 2019 18:52:53 +0300 Subject: RDMA/rw: Use IB_WR_REG_MR_INTEGRITY for PI handover Replace the old signature handover API with the new one. The new API simplifes PI handover code complexity for ULPs and improve performance. For RW API it will reduce the maximum number of work requests per task and the need of dealing with multiple MRs (and their registrations and invalidations) per task. All the mappings and registration of the data and the protection buffers is done by the LLD using a single WR and a special MR type (IB_MR_TYPE_INTEGRITY) for the PI handover operation. The setup of the tested benchmark (using iSER ULP): - 2 servers with 24 cores (1 initiator and 1 target) - ConnectX-4/ConnectX-5 adapters - 24 target sessions with 1 LUN each - ramdisk backstore - PI active Performance results running fio (24 jobs, 128 iodepth) using write_generate=1 and read_verify=1 (w/w.o patch): bs IOPS(read) IOPS(write) ---- ---------- ---------- 512 1243.3K/1182.3K 1725.1K/1680.2K 4k 571233/528835 743293/748259 32k 72388/71086 71789/93573 Using write_generate=0 and read_verify=0 (w/w.o patch): bs IOPS(read) IOPS(write) ---- ---------- ---------- 512 1572.1K/1427.2K 1823.5K/1724.3K 4k 921992/916194 753772/768267 32k 75052/73960 73180/95484 There is a performance degradation when writing big block sizes. Degradation is caused by the complexity of combining multiple indirections and perform RDMA READ operation from it. This will be fixed in the following patches by reducing the indirections if possible. Signed-off-by: Israel Rukshin Reviewed-by: Max Gurtovoy Reviewed-by: Christoph Hellwig Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/isert/ib_isert.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/isert/ib_isert.c b/drivers/infiniband/ulp/isert/ib_isert.c index 53bc505f5292..4b4998e888b9 100644 --- a/drivers/infiniband/ulp/isert/ib_isert.c +++ b/drivers/infiniband/ulp/isert/ib_isert.c @@ -1677,7 +1677,7 @@ isert_rdma_write_done(struct ib_cq *cq, struct ib_wc *wc) isert_dbg("Cmd %p\n", isert_cmd); - ret = isert_check_pi_status(cmd, isert_cmd->rw.sig->sig_mr); + ret = isert_check_pi_status(cmd, isert_cmd->rw.reg->mr); isert_rdma_rw_ctx_destroy(isert_cmd, isert_conn); if (ret) { @@ -1723,7 +1723,7 @@ isert_rdma_read_done(struct ib_cq *cq, struct ib_wc *wc) iscsit_stop_dataout_timer(cmd); if (isert_prot_cmd(isert_conn, se_cmd)) - ret = isert_check_pi_status(se_cmd, isert_cmd->rw.sig->sig_mr); + ret = isert_check_pi_status(se_cmd, isert_cmd->rw.reg->mr); isert_rdma_rw_ctx_destroy(isert_cmd, isert_conn); cmd->write_data_done = 0; -- cgit v1.2.3 From 97162a1ee8a1735fc7a7159fe08de966d88354ce Mon Sep 17 00:00:00 2001 From: Mauro Carvalho Chehab Date: Sat, 8 Jun 2019 23:27:03 -0300 Subject: docs: infiniband: convert docs to ReST and rename to *.rst The InfiniBand docs are plain text with no markups. So, all we needed to do were to add the title markups and some markup sequences in order to properly parse tables, lists and literal blocks. At its new index.rst, let's add a :orphan: while this is not linked to the main index.rst file, in order to avoid build warnings. Signed-off-by: Mauro Carvalho Chehab Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/ipoib/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/Kconfig b/drivers/infiniband/ulp/ipoib/Kconfig index cda8eac55fff..569d614d0c41 100644 --- a/drivers/infiniband/ulp/ipoib/Kconfig +++ b/drivers/infiniband/ulp/ipoib/Kconfig @@ -6,7 +6,7 @@ config INFINIBAND_IPOIB transports IP packets over InfiniBand so you can use your IB device as a fancy NIC. - See Documentation/infiniband/ipoib.txt for more information + See Documentation/infiniband/ipoib.rst for more information config INFINIBAND_IPOIB_CM bool "IP-over-InfiniBand Connected Mode support" -- cgit v1.2.3 From 5d7d78eaecc5c91c9b2001e544a0ae2788d40d1c Mon Sep 17 00:00:00 2001 From: Fuqian Huang Date: Fri, 28 Jun 2019 01:38:04 +0800 Subject: IB/ipoib: Remove memset after vzalloc in ipoib_cm.c vzalloc has already zeroed the memory. So a memset is unneeded. Signed-off-by: Fuqian Huang Reviewed-by: Jason Gunthorpe Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/ipoib/ipoib_cm.c | 1 - 1 file changed, 1 deletion(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_cm.c b/drivers/infiniband/ulp/ipoib/ipoib_cm.c index aa9dcfc36cd3..c59e00a0881f 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_cm.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_cm.c @@ -1153,7 +1153,6 @@ static int ipoib_cm_tx_init(struct ipoib_cm_tx *p, u32 qpn, ret = -ENOMEM; goto err_tx; } - memset(p->tx_ring, 0, ipoib_sendq_size * sizeof(*p->tx_ring)); p->qp = ipoib_cm_create_tx_qp(p->dev, p); memalloc_noio_restore(noio_flag); -- cgit v1.2.3 From 91b01061fef9c57d2f5b712a6322ef51061f4efd Mon Sep 17 00:00:00 2001 From: Valentine Fatiev Date: Sun, 30 Jun 2019 16:48:41 +0300 Subject: IB/ipoib: Add child to parent list only if device initialized Despite failure in ipoib_dev_init() we continue with initialization flow and creation of child device. It causes to the situation where this child device is added too early to parent device list. Change the logic, so in case of failure we properly return error from ipoib_dev_init() and add child only in success path. Fixes: eaeb39842508 ("IB/ipoib: Move init code to ndo_init") Signed-off-by: Valentine Fatiev Reviewed-by: Feras Daoud Signed-off-by: Leon Romanovsky Reviewed-by: Jason Gunthorpe Signed-off-by: Jason Gunthorpe --- drivers/infiniband/ulp/ipoib/ipoib_main.c | 34 ++++++++++++++++++------------- 1 file changed, 20 insertions(+), 14 deletions(-) (limited to 'drivers/infiniband/ulp') diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c index 9b5e11d3fb85..bb904ec511be 100644 --- a/drivers/infiniband/ulp/ipoib/ipoib_main.c +++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c @@ -1893,12 +1893,6 @@ static void ipoib_child_init(struct net_device *ndev) struct ipoib_dev_priv *priv = ipoib_priv(ndev); struct ipoib_dev_priv *ppriv = ipoib_priv(priv->parent); - dev_hold(priv->parent); - - down_write(&ppriv->vlan_rwsem); - list_add_tail(&priv->list, &ppriv->child_intfs); - up_write(&ppriv->vlan_rwsem); - priv->max_ib_mtu = ppriv->max_ib_mtu; set_bit(IPOIB_FLAG_SUBINTERFACE, &priv->flags); memcpy(priv->dev->dev_addr, ppriv->dev->dev_addr, INFINIBAND_ALEN); @@ -1941,6 +1935,17 @@ static int ipoib_ndo_init(struct net_device *ndev) if (rc) { pr_warn("%s: failed to initialize device: %s port %d (ret = %d)\n", priv->ca->name, priv->dev->name, priv->port, rc); + return rc; + } + + if (priv->parent) { + struct ipoib_dev_priv *ppriv = ipoib_priv(priv->parent); + + dev_hold(priv->parent); + + down_write(&ppriv->vlan_rwsem); + list_add_tail(&priv->list, &ppriv->child_intfs); + up_write(&ppriv->vlan_rwsem); } return 0; @@ -1958,6 +1963,14 @@ static void ipoib_ndo_uninit(struct net_device *dev) */ WARN_ON(!list_empty(&priv->child_intfs)); + if (priv->parent) { + struct ipoib_dev_priv *ppriv = ipoib_priv(priv->parent); + + down_write(&ppriv->vlan_rwsem); + list_del(&priv->list); + up_write(&ppriv->vlan_rwsem); + } + ipoib_neigh_hash_uninit(dev); ipoib_ib_dev_cleanup(dev); @@ -1969,15 +1982,8 @@ static void ipoib_ndo_uninit(struct net_device *dev) priv->wq = NULL; } - if (priv->parent) { - struct ipoib_dev_priv *ppriv = ipoib_priv(priv->parent); - - down_write(&ppriv->vlan_rwsem); - list_del(&priv->list); - up_write(&ppriv->vlan_rwsem); - + if (priv->parent) dev_put(priv->parent); - } } static int ipoib_set_vf_link_state(struct net_device *dev, int vf, int link_state) -- cgit v1.2.3